DISCRETE CHANNEL SIMULATION OF BLUETOOTH PICONETS

Document Sample
DISCRETE CHANNEL SIMULATION OF BLUETOOTH PICONETS Powered By Docstoc
					DISCRETE CHANNEL SIMULATION OF BLUETOOTH PICONETS
Workshop on Broadband Wireless Ad-Hoc Networks and Services
12th - 13th September 2002, ETSI, Sophia Antipolis, France

Beatriz Bardón Rodríguez Matilde P. Sánchez Fernández Ana García Armada Department of Signal Theory and Communications Carlos III University of Madrid, Spain e-mail: {beatriz, mati, agarcia}@tsc.uc3m.es

Introduction and motivation
 After the success of Wireless Local Area Networks (WLAN), Bluetooth has come out as an initiative to build Wireless Personal Area Network (WPAN) systems:
 Idea: To connect every device that we are used to carry with us (cellular phones, PDAs, laptops, printers, …)

 The success of Bluetooth depends on the massive use of the standard
 Development of applications that respond to the user’s needs

 Bluetooth devices transmit in ISM band
Worldwide availability of frequencies

Bands also used by many other devices

Rapid introduction to the market

Degradation in throughput and quality
2

Carlos III University of Madrid. Dept. Signal Theory and Communications

Introduction and motivation (II)
 The coexistence of a high number of devices in the same frequency band has a great impact in the applications

 Design of Applications - two options:
 Conservative to ensure a quick market introduction  Optimum in the sense of being capable of making fuller use of the possibilities of the communication

 Simulation techniques
 To analyse the different choices for new services  To characterise in detail every significant effect that influences the system performance
Carlos III University of Madrid. Dept. Signal Theory and Communications

3

Introduction and motivation (III)
 To maintain a good characterisation of the system usually implies long simulation runs  The low bit error rates involved in the case of inclusion of some kind of channel coding imply a great computational cost in simulation  Simulations must be efficient in order to be feasible

DISCRETE CHANNEL MODELS  Useful tool for simulating communication systems operating over fading channels  The bursty nature of errors generated is reproduced by means of a state diagram that avoids the simulation of the whole physical channel

Carlos III University of Madrid. Dept. Signal Theory and Communications

4

Outline
 Bluetooth
 Interference Inmunity and Multiple Access Scheme  Ad Hoc Networks

 Discrete Channel Models for Wireless Communications
 Parameters of a Markov Model  Estimating the parameters of the HMM

 Discrete channel simulation of Bluetooth piconets  Conclusions

Carlos III University of Madrid. Dept. Signal Theory and Communications

5

Bluetooth

 Provides ad-hoc connections via radio using portable devices characterized by
 Low cost  Small size  Low power comsumption
0 dBm for most applications Specifications allow to transmit up to 20 dBm

 This wireless technology must support both voice and data to be transmitted over a short range distance (up to 10 meters typically)
Carlos III University of Madrid. Dept. Signal Theory and Communications

6

Bluetooth: Interference Inmunity and Multiple Access Scheme
 Bluetooth uses the ISM band Interferences coming from other devices (microwave ovens, WLANs) and other Bluetooth devices  To obtain the desired interference inmunity Two options
 Interference suppression  Interference avoidance DSSS FHSS

 Multiple Access Scheme
 FH-CDMA (79 separate 1 MHz channels)  TDD (Time Division Duplex)

A

B

Carlos III University of Madrid. Dept. Signal Theory and Communications

7

Bluetooth: Ad Hoc Networks
 No difference between radio units (Peer communications)  One unit has the ‘master’ role governing the synchronization of the FH communication
A master and one or several slaves (8 max) PICONET
M S

Two or more piconets overlapped in time and space

S

M
S S S M

M S S

S S

SCATTERNET

Carlos III University of Madrid. Dept. Signal Theory and Communications

8

Bluetooth: system block diagram
SCRAMBLING

FRAME CONFORMING

CODING

(whitening)

72 bits

ACCESS CODE

0-2745 bits FREQUENCY HOPPING HEADER PAYLOAD

54 bits

GFSK MODULATOR

Radio Channel 1/3 •FEC
•Multipath •Interferences (piconets, scatternets, other devices, ...)

•NO CODE
GFSK DEMODULATOR

FREQUENCY 1/3 0.5 HOPPING •FECBT = •FEC0.28<h<0.35 2/3

FRAME EXTRACTING

SCRAMBLING
(whitening)
Carlos III University of Madrid. Dept. Signal Theory and Communications

DECODING

9

Discrete Channel Models for Wireless Communications
 Idea: To reproduce, by means of a state diagram, the bursty nature of errors generated in that type of channels  In order to obtain efficient models the simulation is structured in several levels

b

Applications
voice,symbols multimedia, ...
b’

Interleaver, encoder,...

a a a

DCM 2 symbols
symbols
Deinterleaver, decoder,...

a’ a’ a’

Filters, Modulator Filters, equalizers, ... Modulator equalizers, ... physical signals (t) physical signalsGenerate symbols sequences with (t) channel channel the same Filters, characteristics bursty Demodulator Filters, ... equalizers, Demodulator equalizers, ...

Discrete DCM 1 Channel Model
LEVEL 1 LEVEL 2

Carlos III University of Madrid. Dept. Signal Theory and Communications

10

Modelling the behaviour of the black box
b

Applications
voice, multimedia, ...
b’

Interleaver, encoder,...

a

Modulator

Filters , equalizers, ...

physical channel
Deinterleaver, decoder,...

a’

Demodulator

equalizers, ...

Filters,

LEVEL 1 LEVEL 2

Channel state

Describes the behaviour of the channel (a-a’) along time

Finite state channel •Good visualize the channel as being in one of a set of We conditions limited and identifiable finite state states •Fading How is the conditions or channel model obtained?
Most cases Simple cases Derived from simulated or measured error the underlying components Derived analytically from the models of patterns between a-a’ •Noise between a-a’ •Jamming, ..,

Hidden Markov Models (HMM)
Carlos III University of Madrid. Dept. Signal Theory and Communications

11

Parameters of a Markov Model
 Set of states {1, 2, ...N}  State at time t: St  Set of state probabilities: P i(t) = probability of being in state i at time t  Set of state transition probabilities: aij(t) = probability of going from state i at time t to state j at time t+1  Set of input to output transition probabilities for each state: bi(ek) = probability of obtaining the error symbol ek when St = i (from the possible ones {e1, e2, ..., eM})

Carlos III University of Madrid. Dept. Signal Theory and Communications

12

Parameters of a Markov Model (II)
Two states

agb
agg

Transition probabilities

abb
0 1
0.9

Input to output

0
0.1

transition probabilities 0.1
0.9

abg

0
0.7

0.3

0
0.7

1

1

0.3

1
bb0 = 0.3

bg0 = 0.9 bg1 = 0.1

Error symbols = { 0, 1 }
Carlos III University of Madrid. Dept. Signal Theory and Communications

bb1 = 0.7
13

Estimating the parameters of the HMM
 Under certain conditions all the parameters can be inferred from the ... a a  estimation of two matrices
State transition matrix
 . A    . a N1 
11

...

.   .  a NN  

1M

Input to output transition probabilities matrix

 b11  . B    . b N 1 

...

...

b1 M  .   .  b NM  

 Problem to solve Estimate A and B  Data to start Error sequence obtained from the lowest level of simulation  Tool to use Well-known iterative procedure Baum-Welch algorithm  A pair of matrices A, B that generate error sequences with the same characteristics that the one used to train the algorithms are obtained
Carlos III University of Madrid. Dept. Signal Theory and Communications

14

Discrete channel simulation of Bluetooth piconets
FRAME CONFORMING SCRAMBLING (whitening) CODING GFSK MOD

Radio Channel

Error sequence to train the algorithm
•Multipath (Channel models for HIPERLAN/2 in different indoor scenarios) •Interferences ( scatternets, microwave ovens, ...)

1 =Simulationsdecision Erroneous with

0 =different decision Correct coding schemes
GFSK signal

FRAME EXTRACTING

SCRAMBLING

DECODING

GFSK DEMOD

(whitening)

Carlos III University of Madrid. Dept. Signal Theory and Communications

15

Discrete channel simulation of Bluetooth piconets (II)
 The direct application of the Baum-Welch algorithm requires great amount of computations, specially when the error sequence contains long chains of identical symbols  K. S. Shanmugan et al. have proposed a modified version of the BW algorithm that involves great saving in computation  Once the parameters (A,B matrices) have been obtained it is indispensable to validate them for ensuring the use of the discrete channel model in upper levels of simulation

 Comparison between the error sequence arising from the original physical layer and the one generated by our HMM
 Cross-correlation between the two sequences  Histograms characterising error free intervals for the different guard times are obtained and compared (chi-square goodnes-of-fit test)
Carlos III University of Madrid. Dept. Signal Theory and Communications

16

Conclusions
 A Discrete Channel simulation method for the efficient evaluation of Bluetooth radio system has been proposed .  Structuring the simulation in several levels and modelling them by means of Hidden Markov Models allows a great saving in computational resources.  It will be very useful to obtain models for different design options and environments.
Need for standardisation: It would be very convenient to have these Discrete Channel Models standardised for WPAN (as it has been done with GSM) in order to be able to evaluate the performance of new applications.
Carlos III University of Madrid. Dept. Signal Theory and Communications

17


				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:4
posted:9/14/2009
language:English
pages:17
Lingjuan Ma Lingjuan Ma
About